Folliculin encoded by the BHD gene interacts with abinding protein, FNIP1, and AMPK, and is involvedin AMPK and mTOR signalingMasaya Baba*†, Seung-Beom Hong*†, Nirmala Sharma*, Michelle B. Warren*†, Michael L. Nickerson*‡,Akihiro Iwamatsu§, Dominic Esposito¶, William K. Gillette¶, Ralph F. Hopkins III¶, James L. Hartley¶, Mutsuo Furihata�,Shinya Oishi**, Wei Zhen*, Terrence R. Burke, Jr.**, W. Marston Linehan†, Laura S. Schmidt*†,††‡‡, and Berton Zbar*

Laboratories of *Immunobiology and **Medicinal Chemistry, Center for Cancer Research, National Cancer Institute–Frederick, Frederick, MD 21702; ¶ProteinExpression Laboratory, Research Technology Program and ††Basic Research Program, SAIC–Frederick, Inc., National Cancer Institute–Frederick, Frederick, MD21702; �Department of Pathology, Kochi Medical School, Kochi University, Kochi 783-8505, Japan; §Protein Research Network, Yokohama 236-0004, Japan;and †Urologic Oncology Branch, Center for Cancer Research, National Cancer Institute, National Institutes of Health, Bethesda, MD 20894

Edited by Bert Vogelstein, The Sidney Kimmel Comprehensive Cancer Center at Johns Hopkins, Baltimore, MD, and approved August 23, 2006(received for review May 8, 2006)

Birt–Hogg–Dube syndrome, a hamartoma disorder characterizedby benign tumors of the hair follicle, lung cysts, and renal neopla-sia, is caused by germ-line mutations in the BHD(FLCN) gene, whichencodes a tumor-suppressor protein, folliculin (FLCN), with un-known function. The tumor-suppressor proteins encoded by genesresponsible for several other hamartoma syndromes, LKB1,TSC1�2, and PTEN, have been shown to be involved in the mam-malian target of rapamycin (mTOR) signaling pathway. Here, wereport the identification of the FLCN-interacting protein, FNIP1, anddemonstrate its interaction with 5� AMP-activated protein kinase(AMPK), a key molecule for energy sensing that negatively regu-lates mTOR activity. FNIP1 was phosphorylated by AMPK, and itsphosphorylation was reduced by AMPK inhibitors, which resultedin reduced FNIP1 expression. AMPK inhibitors also reduced FLCNphosphorylation. Moreover, FLCN phosphorylation was dimin-ished by rapamycin and amino acid starvation and facilitated byFNIP1 overexpression, suggesting that FLCN may be regulated bymTOR and AMPK signaling. Our data suggest that FLCN, mutatedin Birt–Hogg–Dube syndrome, and its interacting partner FNIP1may be involved in energy and�or nutrient sensing through theAMPK and mTOR signaling pathways.

hamartoma syndrome � renal cancer � Birt–Hogg–Dube � tumor suppressor

B irt–Hogg–Dube (BHD) syndrome predisposes patients todevelop hair follicle hamartomas, lung cysts, and an in-

creased risk for renal neoplasia (1–3). BHD patients developbilateral, multifocal renal tumors with a variety of histologies (4).We mapped the BHD locus to chromosome 17p11.2 by linkageanalysis in BHD kindreds (5, 6) and identified germ-line muta-tions in a gene with unknown function that is highly conserved(7, 8). Twenty-two unique mutations predicted to truncate theBHD protein folliculin (FLCN), including a ‘‘hot spot’’ inser-tion�deletion in a C8 tract, were identified in 84% of BHDkindreds (9). Somatic ‘‘second-hit’’ mutations identified inBHD-associated renal tumors suggest a tumor-suppressor func-tion for FLCN (10), underscored by loss of BHD mRNAexpression in renal tumors from BHD patients (11).

Recent studies suggest that several hamartoma syndromes maybe linked through the convergent energy�nutrient-sensing path-ways involved in mammalian target of rapamycin (mTOR) regu-lation (12–15). These inherited syndromes are characterized bymultiple hamartomas and an increased risk of cancer. Germ-linemutations have been identified in four causative genes: LKB1,responsible for Peutz–Jeghers syndrome (16–18), TSC1 andTSC2, responsible for tuberous sclerosis complex (TSC) (19), andPTEN, responsible for Cowden syndrome (20). Loss of genefunction leads to dysregulation of mTOR, which regulates cellgrowth and size through stimulation of protein synthesis (15, 21, 22).

BHD syndrome, also a hamartoma disorder, displays phenotypicsimilarities to TSC that have led to speculation that BHD mayfunction in the pathway(s) signaling through mTOR (12, 23). Toascertain FLCN function, we searched for interacting proteins bycoimmunoprecipitation. We identified a 130-kDa FLCN-interacting protein, FNIP1, and demonstrated its interaction withAMPK, a protein important in nutrient�energy sensing (24, 25).We found that FLCN is phosphorylated, which is facilitated byFNIP1, and suppressed by conditions that inhibit mTOR. We alsofound that AMPK-inhibitor treatment reduced FNIP1 phosphor-ylation, lowered FNIP1 expression levels, and resulted in FLCNdephosphorylation. We discuss FLCN–FNIP1–AMPK interactionsand consider how dysregulation of this complex may predispose tothe BHD phenotype.

ResultsIdentification of a 130-kDa FLCN-Interacting Protein, FNIP1, andCloning of Human FNIP1 cDNA. BHD encodes FLCN, a highlyconserved protein with unknown function. To elucidate its biologicrole, we searched for FLCN-interacting protein(s) in anti-HAimmunoprecipitates from doxycycline-inducible, HA-FLCN-expressing HEK293 cells. Mass spectrometric analysis of a major130-kDa interacting protein (Fig. 1A), designated FNIP1 (FLCN-interacting protein 1), identified peptides from two overlappingproteins, BAB85547, encoded by KIAA1961 (AB075841), andAAH01956, encoded by an uncharacterized human EST(BC001956; see peptide mass fingerprinting in Data Set 1, which ispublished as supporting information on the PNAS web site).Primers designed from the overlapping cDNA sequences were usedto amplify and clone the full-length FNIP1 transcript (DQ145719)from a pooled tissue cDNA library. Several alternative transcriptslacking one or two of the 18 coding exons were also identified and

Author contributions: M.B., S.-B.H., W.M.L., L.S.S., and B.Z. designed research; M.B., N.S.,M.B.W., M.L.N., A.I., and M.F., performed research; M.B., J.L.H., D.E., W.K.G., R.F.H., W.Z.,S.O., T.R.B., and W.M.L. contributed new reagents�analytic tools; M.B., S.-B.H., M.L.N., A.I.,L.S.S., W.M.L. and B.Z. analyzed data; and M.B., L.S.S., and B.Z. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS direct submission.

Freely available online through the PNAS open access option.

Abbreviations: BHD, Birt–Hogg–Dube; FLCN, folliculin; mTOR, mammalian target of rapa-mycin.

Data deposition: The sequence reported in this paper has been deposited in the GenBankdatabase (accession no. DQ145719).

‡Present address: Translational and Clinical Research, Transgenomic, Inc., Gaithersburg,MD 20878.

‡‡To whom correspondence should be addressed at: National Cancer Institute–Frederick,Building 560, Room 12-69, Frederick, MD 21702. E-mail: schmidtl@ncifcrf.gov.

© 2006 by The National Academy of Sciences of the USA

15552–15557 � PNAS � October 17, 2006 � vol. 103 � no. 42 www.pnas.org�cgi�doi�10.1073�pnas.0603781103

are being verified (details in Supporting Methods, which is publishedas supporting information on the PNAS web site). The evolution-arily conserved 3,626-nt mRNA encoded a protein with an ORF of1,166 aa. A comparison of FNIP1 homologs from Homo sapiens,Xenopus tropicalis, Danio rerio, Drosophila melanogaster, and Cae-norhabditis elegans revealed five blocks of conserved sequence withat least 35% similarity (Fig. 1C; and see Fig. 8, which is publishedas supporting information on the PNAS web site), but BLASTanalysis revealed no known functional domains. Northern blotanalysis showed that FNIP1 was expressed in a pattern similar toBHD, with strongest expression in heart, liver, and placenta andmoderate expression in tissues involved in the BHD phenotype,kidney and lung (Fig. 1B).

FLCN and FNIP1 Interact in Vivo Through the C Terminus of FLCN andColocalize in the Cytoplasm. To confirm the interaction betweenFNIP1 and FLCN, we performed reciprocal coimmunoprecipita-tions in HEK293 cells transiently expressing HA-FLCN andFLAG-FNIP1 and found that HA-FLCN coimmunoprecipitatedwith FLAG-FNIP1 in a reciprocal fashion (Fig. 2A). Similarly,endogenous FNIP1–FLCN interaction was confirmed in HEK293cells, blotting with polyclonal antibodies to FLCN and FNIP1 (Fig.2B). HA-FNIP1 and FLAG-FLCN, transiently cotransfected intoHeLa cells, colocalized in the cytoplasm with a reticular pattern(Fig. 2C). FLCN also localized in the nucleus.

To identify the FNIP1-binding domain in FLCN, in vitro bindingassays were performed with GST-FLCN mutants and 35S in vitro-translated FNIP1. FNIP1 bound to full-length GST-FLCN (1–579aa) and mutant proteins containing the C terminus of FLCN(246–579 or 345–579 aa; Fig. 3A) but not to mutant proteins lackingthe C terminus (1–516 or 1–344 aa). These interactions wereconfirmed in immunoprecipitates from HA-FNIP1-expressingHEK293 cells transiently transfected with FLAG-FLCN andFLAG-FLCN mutants (Fig. 9, which is published as supportinginformation on the PNAS web site). In support of these data, themutant protein from the most C-terminal BHD mutation found inpatients (c.2034C�T; R527X) (9) did not bind to HA-FNIP1 (datanot shown). Taken together, the results support a FNIP1-bindingdomain in the C terminus of FLCN.

Optimal Binding of FLCN to FNIP1 Requires Full-Length FNIP1 Protein.We next performed an in vitro binding assay to determine theportion of FNIP1 that was required to bind FLCN. Full-lengthFNIP1 protein was required for maximum binding of FLCN; onlyone of the deletion mutants containing residues 300-1166 (lackingconserved block 1, see Fig. 1C) bound FLCN (Fig. 3B). BothGST-FNIP1 and the FNIP1 mutant protein that bound FLCNcontained four of the five conserved blocks (blocks 2–5). Mutantscontaining either blocks 2 and 3 or blocks 4 and 5 did not bindFLCN. The data suggest that optimal FLCN–FNIP1 interactionrequires these four conserved regions of the FNIP1 protein, per-haps for proper conformation.

A B

C

Fig. 1. FNIP1, a 130-kDa FLCN-binding protein, is evolutionarily conservedand widely expressed in human tissues. (A) Stably transfected HEK293 cellsexpressing doxycycline-inducible HA-FLCN were lysed, and proteins wereimmunoprecipitated with anti-HA antibody, subjected to SDS�PAGE, trans-ferred to PVDF membranes, and stained with colloidal gold. FNIP1 was coim-munoprecipitated with FLCN. (B) FNIP1 and BHD(FLCN) expression patterns inhuman adult tissues. cDNA probes were generated by PCR and hybridized toa human 12-tissue Northern blot (Clontech). �-Actin was used as loadingcontrol. (C) Schematic structure of conserved FNIP1 domains. Amino acidalignments show five blocks of conservation between FNIP1 and its homologs(Fig.8). GenBank accession nos. for FNIP1 homologs: H. sapiens (DQ145719), X.tropicalis (ENSXETP00000036209; Ensembl), D. rerio (XM�687716), D. melano-gaster (NM�140686), and C. elegans (T04C4.1a; Wormbase).

A B

C

Fig. 2. FLCN–FNIP1 interactions. (A) HA-FLCN was cotransfected with FLAG-FNIP1 into HEK293 cells. Cell lysates were immunoprecipitated (IP) with anti-FLAG or anti-HA antibodies and blotted with both antibodies. Arrows, FLAG-FNIP1; arrowheads, HA-FLCN. (B) HEK293 lysates were immunoprecipitatedwith anti-FLCN and control IgG or with anti-FNIP1 and control IgG and blottedwith anti-FLCN or anti-FNIP1. Arrowheads, endogenous FLCN; arrows, endog-enous FNIP1; asterisks, IgG heavy-chain bands. (C) HeLa cells were cotrans-fected with HA-FNIP1 and FLAG-FLCN. Immunocytochemistry was performedwith anti-HA and anti-FLAG antibodies. FNIP1 and FLCN colocalize with areticular pattern in the cytoplasm. Green, FLAG-FLCN; red, HA-FNIP1.

A

B

Fig. 3. Regions of FNIP1 and FLCN necessary for binding. (A) RecombinantGST-FLCN and GST-FLCN mutants were immobilized on glutathione-Sepharose and incubated with 35S in vitro-translated (IVT) FNIP1. Binding wasevaluated by SDS�PAGE and autoradiography. Coomaissee Brillant blue (CBB)staining shows relative expression of GST-FLCN proteins. FLCN mutants lackinga C terminus cannot bind FNIP1. (B) Recombinant GST-FNIP1 and FNIP1 dele-tion mutants were bound to glutathione-Sepharose, incubated with 35S IVTFLCN, and evaluated as in A. Deletion of conserved blocks 2�3 or 4�5 of FNIP1disrupted the interaction with FLCN.

Baba et al. PNAS � October 17, 2006 � vol. 103 � no. 42 � 15553

MED

ICA

LSC

IEN

CES

AMPK Was Identified as a FNIP1-Interacting Protein. Because FNIP1is a protein with no characteristic domains to suggest function, wesearched for interacting proteins using the experimental approachdescribed above. We identified the �-1 subunit (40 kDa) of AMPKas well as FLCN (67 kDa) and FNIP1 in the immunoprecipitatesfrom doxy-inducible, HA-FNIP1-expressing HEK293 cells by massspectrometry (Fig. 4A; peptide mass fingerprinting in Data Sets2–4, which are published as supporting information on the PNASweb site). Additional bands included HSP90 (p90) and severalunverified proteins. AMPK is known to play a critical role in energysensing by negatively regulating biosynthetic pathways and activat-ing catabolic reactions in response to stress signals (24–26). Thisheterotrimeric protein kinase consists of a catalytic �-subunit andtwo regulatory subunits, � and �. Anti-HA immunoprecipitatesfrom doxy-inducible HA-FNIP1-expressing HEK293 cells wereWestern blotted with subunit-specific antibodies and found tocontain both endogenous AMPK �- and �1 subunits (Fig. 4B). Weconfirmed the interaction of endogenous FNIP1 and AMPK� inanti-FNIP1 immunoprecipitates by Western blotting (Fig. 4D),confirming FNIP1 interaction with all subunits of AMPK. AMPKbinding to FNIP1 was confirmed by an in vitro binding assayincubating HEK293 cell lysates (AMPK source) with recombinantGST-FNIP1 immobilized on glutathione-Sepharose, followed byWestern analysis. Residues 602–929 were the minimum region ofFNIP1 necessary for binding to AMPK, which was distinct from theFLCN-binding region. Direct binding of recombinant GST-FNIP1to purified AMPK was also demonstrated in vitro (Fig.10 A and B,

which is published as supporting information on the PNAS website).

The catalytic �-subunit of AMPK is activated by phosphorylationon Thr-172 by AMPK kinases (27). We observed that the amountof AMPK� present in its phosphorylated form (detected by aphospho-AMPK� Thr-172-specific antibody) relative to totalAMPK� was greater in HA-FNIP1 immunoprecipitates than intotal cell lysates (Fig. 4C Upper). Using densitometry, we comparedthe total AMPK� and phospho-AMPK� in doxy-induced HA-FNIP1-expressing lysates with the amounts of these proteins inanti-HA immunoprecipitates from the lysates. We estimated that5% of the total phospho-AMPK� but only 0.5% of the totalAMPK� associated with FNIP1 in the immunoprecipitates (Fig. 4CLower).

FLCN Binds to AMPK Through FNIP1 but Is Not Essential for FNIP1–AMPK Interaction. To determine whether a FLCN�FNIP1�AMPKcomplex exists, we cotransfected HEK293 cells with an HA-FLCNexpression vector with or without a FLAG-FNIP1 vector. Anti-HAimmunoprecipitates were evaluated by Western blotting. All threeAMPK subunits were coimmunoprecipitated with HA-FLCN in aFLAG-FNIP1-dependent manner (Fig. 4E, lane 4) confirming thatthe AMPK heterotrimer binds to the FLCN–FNIP1 complex in aFNIP1-dependent manner but not to FLCN alone (Fig. 4E, lane 3).We also evaluated the FNIP1–AMPK interaction in a FLCN-nullrenal tumor cell line, UOK257, and in UOK257-2 cells retrovirallyrestored with FLCN. In UOK257 cells, no FLCN was detected intotal-cell lysates or anti-FNIP1 immunoprecipitates (Fig. 4F, lanes1 and 3). However, endogenous FNIP1 was able to bind to theAMPK� subunit in anti-FNIP1 immunoprecipitates (Fig. 4F, lane3), and AMPK–FNIP1 binding was not affected by restoration ofFLCN expression (UOK257-2; Fig. 4F, lanes 4 and 6).

FNIP1 Can Be Phosphorylated by AMPK in the FLCN�FNIP1�AMPKComplex, Which Affects Its Expression Level. To determine whetherAMPK in the FNIP1 immunoprecipitates was enzymatically active,we performed an in vitro kinase assay with anti-HA immunopre-cipitates from doxycycline-induced HA-FNIP1-expressingHEK293 cells using SAMS peptide, a specific AMPK substrate(28), and [�-32P]ATP. Immunoprecipitates from doxycycline(�)lysates showed significant AMPK activity compared with controls.The FNIP1-associated in vitro kinase activity was increased 2.5-foldby 75 �M AMP, a positive effector of AMPK (26) (Fig. 5A),supporting AMPK as the source of kinase activity.

We next asked whether FNIP1 might serve as a substrate forAMPK in the FNIP1–AMPK complex. When anti-HA immuno-precipitates from doxycycline-induced HA-FNIP1-expressingHEK293 cells were incubated with [�-32P]ATP and AMPK kinasebuffer, we found that HA-FNIP1 was phosphorylated, and thisphosphorylation was blocked by Compound C, a specific inhibitorof AMPK (29), in a dose-dependent manner (Fig. 5B), whereasother kinase inhibitors tested (i.e., UO126, Wortmannin, andrapamycin) were unable to block HA-FNIP1 phosphorylation.Furthermore, when we labeled doxy-induced HA-FNIP1-expressing HEK293 cells in vivo with [32P]orthophosphate for 3 h,addition of Compound C or AraA (another inhibitor of AMPK)(30) to the medium reduced the amount of [32P]phosphate incor-porated into HA-FNIP1 (Fig. 5C), indicating that FNIP1 may serveas a substrate for AMPK both in vitro and in vivo.

We postulated that one function of FNIP1 phosphorylation byAMPK might be regulation of protein expression. We cultured bothHA-FNIP1-overexpressing HEK293 cells and 293 cells in thepresence of AMPK inhibitors for 24 h and found that levels ofHA-FNIP1 and endogenous FNIP1 proteins were reduced (Fig. 5D and E, respectively). Levels of endogenous FLCN were alsoreduced by AMPK inhibition.

A B C

D E F

Fig. 4. FNIP1 interacts with AMPK in vitro and in vivo. (A) FNIP1-interactingproteins were immunoprecipitated (IP) with anti-HA antibody from doxycy-cline-inducible HA-FNIP1-expressing HEK293 cells, separated by SDS�PAGE,transferred to PVDF membrane, stained with colloidal gold, and analyzed bymass spectrometry. The 40-kDa protein was identified as the �-1 subunit ofAMPK. Other interacting proteins were FLCN (67 kDa) and HSP 90 (90 kDa). (B)Cell lysates from HA-FNIP1-inducible HEK 293 cells cultured with and withoutdoxy were immunoprecipitated with anti-HA antibody and blotted withantibodies, as indicated, showing that HA-FNIP1 and AMPK subunits interact.(C) Immunoprecipitates from HA-FNIP1-inducible HEK293 cells were blottedwith antibodies as indicated. Band intensities were measured with the ImageJprogram. The estimated amounts of AMPK� and phospho-AMPK�(T172)binding to FNIP1 were calculated by IP�[input (4%) � 25]. Results were fromthree independent experiments. Brackets, standard deviation. (D) HEK293lysates were immunoprecipitated with anti-FNIP1 or control IgG. Immunopre-cipitates were blotted with anti-FNIP1, anti-FLCN mouse monoclonal anti-body, and anti-AMPK� mouse monoclonal antibody. Endogenous FNIP1 andAMPK interact. (E) HEK293 cell lysates transfected (T.F.) with HA-FLCN aloneor HA-FLCN and FLAG-FNIP1,were immunoprecipitated with anti-HA anti-body and blotted with antibodies as indicated. All AMPK subunits wereimmunoprecipitated with FLCN in a FNIP1-dependent manner. (F) UOK257cells (FLCN null) and FLCN-restored UOK257-2 cells were immunoprecipitatedwith anti-FNIP1 or control IgG and blotted with anti-FNIP1, anti-FLCNmAb,and anti-AMPK� mAb. W.B., Western blotting; Sup, supernatant.

15554 � www.pnas.org�cgi�doi�10.1073�pnas.0603781103 Baba et al.

FLCN Can Exist in a Phosphorylated Form That Is Preferentially Boundto FNIP1. In searching for a functional significance for the FLCN–FNIP1–AMPK interaction, we noted the appearance of multipleFLCN bands on Western blots and designed experiments to testtheir association with or regulation by FNIP1. Electrophoreticmobility shifts observed after phosphatase or calyculin A (phos-phatase inhibitor) treatment suggested that FLCN existed in severalphosphorylated forms (Fig. 11, which is published as supportinginformation on the PNAS web site). We also detected at least threeelectrophoretically distinct forms of endogenous FLCN in HEK293cell lysates, which shifted to the slower migrating (phosphorylated)species in anti-FNIP1 immunoprecipitates (Fig. 4D), suggestingthat FNIP1 preferentially bound phosphorylated FLCN. In addi-tion, doxy-induced HA-FNIP1 expression produced a band shift tophospho-FLCN in total HEK293 cell lysates (Fig. 4B, lane 2), andthe slower migrating phospho-FLCN preferentially immunopre-cipitated with HA-FNIP1 from doxy-induced HEK293 cells (Fig.4B, lane 4).

FLCN Phosphorylation Is Diminished by Inhibition of mTOR and AMPK.To elucidate the mechanism by which FLCN becomes phosphor-ylated, UOK257-2 FLCN-restored cells were grown under differentculture conditions or with different kinase inhibitors. Serum with-drawal for 48 h partially decreased FLCN phosphorylation (Fig. 6A,lane 1), which was restored by serum stimulation (Fig. 6A, lane 2).Pretreatment with UO126, an inhibitor of MEK1�2, or Wortman-nin, an inhibitor of PI3K, reduced FLCN phosphorylation to someextent (Fig. 6A, lanes3�4 and 2); interestingly, rapamycin, whichcompletely inhibited mTOR activity (no pS6R), most dramaticallyinhibited FLCN phosphorylation by eliminating the upper phos-pho-FLCN band completely (Fig. 6A, lanes 5 and 2). However,rapamycin did not completely block FLCN phosphorylation, be-cause one phospho-FLCN band was unaffected by rapamycintreatment (Fig. 6B, lane 2). Treatment of UOK257-2 with Com-pound C resulted in complete loss of phospho-FLCN, including therapamycin-insensitive phospho-FLCN band (Fig. 6B, lane3) andresulted in reduced FLCN expression. The same results wereobtained with AMPK inhibitor AraA (Fig. 5 D and E). In support

of these data, we demonstrated phosphorylation of recombinantGST-FLCN by purified AMPK in an in vitro kinase reaction (Fig.12, which is published as supporting information on the PNAS website). Taken together, these data indicate that FLCN has multiplephosphorylation sites that are phosphorylated through mTOR andAMPK signaling.

FLCN Phosphorylation, Which Is Facilitated by FNIP1 Expression, IsBlocked by Inhibition of mTOR. As mentioned above, FNIP1 over-expression facilitates FLCN phosphorylation (Figs. 4B and 6C,lanes 1 and 2). Interestingly, rapamycin partially inhibited FLCNphosphorylation facilitated by FNIP1 overexpression in doxycy-cline-induced HA-FNIP1-expressing HEK293 cells (Fig. 6C, lanes2 and 4). Likewise, amino acid starvation, known to inhibit mTORsignaling (see pS6R readout), suppressed FLCN phosphorylationfacilitated by FNIP1 overexpression (Fig. 6C, lanes 2 and 6). Aminoacid stimulation restored FNIP1 overexpression-stimulated FLCNphosphorylation to amino acid-starved cells (Fig. 6C, lanes 6 and 8),and this effect was blocked by rapamycin pretreatment (Fig. 6C,lanes 8 and 10). These data suggest a role for mTOR signaling inmodulating FLCN phosphorylation facilitated by FNIP1 overex-pression, although it is not yet clear whether FLCN is a directsubstrate of mTOR or a substrate of a downstream kinase that isactivated by mTOR. Note that FNIP1 overexpression can stillfacilitate rapamycin-insensitive FLCN phosphorylation (lowestFLCN band shifts up, Fig. 6C, lanes 8 and 10).

Activation of mTOR Signaling by Serum Stimulation and Its Inhibitionby AMPK Activation Is Independent of FLCN Expression, but Responseto Some Cell Stress Conditions Is FLCN-Dependent. The regulation ofFLCN phosphorylation by mTOR and the interaction of AMPKwith FNIP1�FLCN suggested a functional relationship betweenFLCN�FNIP1 and mTOR. We compared the responses of theFLCN-null UOK257 and FLCN-restored UOK257-2 cell lines tophysiological conditions that affect mTOR activity. We found that

A B D

CE

Fig. 5. AMPK phosphorylates FNIP1 and may regulate FNIP1 expression. (A)Cell lysates from HA-FNIP1-inducible HEK293 cells grown with and withoutdoxy were immunoprecipitated with anti-HA antibody and assayed for AMPKin vitro kinase activity with [�-32P]ATP and SAMS peptide with and withoutAMP. Radioactivity incorporated into SAMS peptide was measured by scintil-lation counting. Each bar represents the results of three independent exper-iments. Brackets, standard deviation. (B) HA-FNIP1 immunoprecipitates wereincubated in AMPK kinase reaction buffer with [�32]ATP and various kinaseinhibitors (Compound C, 4 or 40 �M; UO126, 10 �M; Wortmannin, 1 �M; andrapamycin, 20 nM). Immunoprecipitates were washed and subjected to SDS�PAGE and autoradiography. (C) HA-FNIP1-inducible 293 cells were cultured for3 h with [32P]orthophosphate with and without AMPK inhibitors (AraA, 2 mMand Compound C, 20 �M), and anti-HA immunoprecipitates were subjected toSDS�PAGE and autoradiography. Cold samples obtained under the sameculture conditions were used for phosphoacetyl CoA carboxylase (pACC) andHA-FNIP1 Western blotting. (D and E) HA-FNIP1-inducible 293 cells (D) or 293cells (E) were cultured with AraA (2 mM) or Compound C (Comp C) (30 �M) for24 h. Exogenous or endogenous FNIP1 and endogenous FLCN were detectedby Western blotting. WB, Western blotting.

A

C

B

Fig. 6. FLCN phosphorylation is regulated by FNIP1 through mTOR signalingand AMPK. (A) FLCN phosphorylation is affected by mTOR activity. UOK257-2FLCN-restored cells were cultured under different culture conditions: S(�), withserum; S(�), serum starvation for 48 h; or S(�) 3 S(�), 20% dialyzed serumstimulation for 30 min after 48-h serum starvation. Cells were pretreated for 2 hbefore stimulation with various inhibitors: nontreatment (N�T); U0126, 50 �M;Wortmannin (Wort), 1 �M; or rapamycin (Rapa), 20 nM. (B) AMPK inhibitionsuppresses FLCN phosphorylation. UOK 257–2 cells were cultured with rapamycin(20 nM), Compound C (CompC) (30 �M), or both for 24 h. Western blotting (WB)was performed for phospho-ACC (AMPK readout) and phospho-S6R (mTORreadout). Compound C treatment reduced FNIP1 expression and FLCN phosphor-ylation. (C) FLCN phosphorylation facilitated by FNIP1 expression is partiallyregulated by mTOR activity. HA-FNIP1-inducible HEK293 cells were cultured withand without doxy. For amino acid starvation, cells were cultured with Earl’sbalanced solution containing dialyzed serum, vitamins, pyruvate, and glucose for16 h with and without doxycycline, followed by stimulation with 2� amino acidsfor 30 min. Rapamycin was added 2 h before stimulation.

Baba et al. PNAS � October 17, 2006 � vol. 103 � no. 42 � 15555

MED

ICA

LSC

IEN

CES

2-deoxyglucose treatment activated AMPK (increased phospho-ACC), which, in turn, down-regulated mTOR in a comparablefashion in FLCN-null cells and FLCN-restored cells (Fig. 7A).mTOR signaling through PI3K and pAKT under serum-stimulatedconditions and rapamycin inhibition of mTOR signaling, wereunaffected by lack of FLCN expression (Fig. 7B), suggesting thatFLCN may function downstream of the AMPK�mTOR signalingreadouts evaluated in these experiments. On the other hand,mTOR inhibition as a result of serum starvation was significant inFLCN-restored cells but retained some signaling in FLCN-null cells(Fig. 7B). Amino acid deprivation appeared to have the oppositeeffect, shutting down mTOR signaling more effectively in FLCN-null cells (Fig. 7B.) Therefore, FLCN-null cells seem to respondsomewhat differently to certain physiologic conditions that inducecell stress and involve the mTOR signaling pathway.

DiscussionHere, we report the identification of the FLCN-interactingprotein, FNIP1, a 130-kDa protein that is moderately conservedamong species, suggesting that it serves an important function inliving cells. FLCN was shown to interact with FNIP1 through itsC terminus in vivo and in vitro. Interestingly, nearly all mutationsin BHD patients are predicted to produce a C-terminally trun-cated FLCN unable to bind FNIP1, which may compromiseFLCN function and lead to the BHD phenotype.

We have identified AMPK as a FNIP1-interacting protein.Mutations in certain genes in the converging LKB1–AMPK–TSC1�2-mTOR and PI3K–AKT–TSC1�2-mTOR energy- and nu-trient-sensing pathways lead to the development of several hamar-toma syndromes. Cellular energy deficit triggers AMPKphosphorylation by LKB1 and calmodulin kinase kinase (31, 32),which negatively regulates mTOR through TSC2 phosphorylation(29, 33, 34), inhibiting protein translation and cell growth. PTENblocks mTOR through inhibition of PI3K-activation of AKT,induced by growth factors that signal through membrane receptors(35–38). Inactivating mutations that disrupt these pathways result indysregulated cell growth and protein synthesis, leading to a hamar-toma ‘‘overgrowth’’ phenotype (Fig. 13, which is published assupporting information on the PNAS web site). Birt–Hogg–Dubesyndrome predisposes to hair follicle hamartomas and renal neo-plasia. The interaction of FLCN with FNIP1 and AMPK suggestsa possible role for FLCN and FNIP1 in the nutrient�energy-sensingpathways involving AMPK and mTOR and may provide a molec-ular mechanism for the BHD phenotype. Our data showing thatFLCN phosphorylation facilitated by FNIP1 is mTOR- andAMPK-activity-dependent and that FNIP1 and FLCN phosphor-ylation and their expression were suppressed by AMPK inhibitionindicate a functional interaction between FLCN�FNIP1 and

mTOR�AMPK signaling pathways. Our data suggest that FLCNhas multiple phosphorylation sites, and FLCN phosphorylationthrough AMPK may be necessary for additional phosphorylationby mTOR signaling. Furthermore, because AMPK inhibition re-sulted in a decreased level of FNIP1 expression, and overexpressionof FNIP1 facilitated both rapamycin-sensitive and -insensitiveFLCN phosphorylation, FNIP1 may be necessary for FLCN phos-phorylation through both mTOR and AMPK, perhaps serving asa scaffold regulator of FLCN.

The biological mechanism explaining how FLCN functions as atumor suppressor is still unknown. AMPK activation by 2-deoxy-glucose can suppress mTOR activity in FLCN-null cells, indicatingthat FLCN is not necessary for mTOR suppression by AMPK.There was no difference in AMPK activity [phospho-ACC andphospho-AMPK�(T172) readouts] between FLCN-null and-restored cells, suggesting that FLCN is not an upstream regulatorof AMPK. Under normal culture conditions or serum-stimulatedconditions, FLCN-null cells displayed mTOR activity equal torestored cells. But FLCN-null cells exhibited different mTORactivity compared with FLCN-restored cells under certain condi-tions, such as serum starvation and amino acid deprivation, whichis difficult to explain without further experimentation. These dif-ferences may be caused by direct or indirect effects, includingpossible feedback loops.

Based on these findings, we hypothesize that FNIP1 and FLCNmight be downstream effectors of mTOR and AMPK and modu-late energy�nutrient-sensing signaling pathways by some as-yet-unknown molecular mechanism. Further clarification of FLCNfunction as a tumor suppressor may contribute to a broaderunderstanding of the molecular mechanisms responsible for otherhamartoma syndromes caused by mTOR dysregulation.

Materials and MethodsCell Lines and Cell Culture. A FLCN-null cell line, UOK257, wasestablished from a clear-cell renal tumor with 17p loss (10) from aBHD patient with a c.1733insC germ-line mutation (7). A FLCN-restored cell line, UOK257-2, was established by infection ofUOK257 with a BHD-expressing lentiviral vector by using theViraPower Lentiviral Expression System (Invitrogen, Carlsbad,CA). HEK293 cells expressing tetracycline-inducible HA-FLCNand FNIP1-HA were established by using the Flp-In T-Rex System(Invitrogen) according to the manufacturer’s protocol.

Identification of FLCN- and FNIP-1-Interacting Proteins. HA-FLCN-or FNIP1-HA-inducible HEK 293 cells were cultured with orwithout 1 �g�ml doxycycline for 36 h, lysed, immunoprecipitatedwith anti-HA affinity matrix, and eluted with HA peptide accord-ing to the manufacturer’s protocol (Roche Applied Science, Indi-anapolis, IN). The eluted proteins were subjected to SDS�PAGE,transferred to PVDF membranes (ProBlott PVDF; Applied Bio-systems, Foster City, CA), stained with Colloidal Gold TotalProtein stain (Bio-Rad, Hercules, CA), excised and prepared asdescribed (39), and analyzed by MALDI-TOF�MS using an Ap-plied Biosystems Voyager-DE�STR (40) (details in SupportingMethods). Peptide mass fingerprinting and protein databasesearches confirmed protein identities (40).

Molecular Cloning of Human FNIP1 cDNA. The full-length FNIP1transcript of 3,626 nt (DQ145719) was amplified from a pooledtissue cDNA library (Clontech, Mountain View, CA) and se-quenced to 6-fold coverage (details in Supporting Methods).K1AA1961 cDNA was obtained from Kazusa DNA ResearchInstitute (Chiba, Japan).

Antibodies. Normal rabbit IgG, HRP-labeled anti-mouse and anti-rabbit secondary antibodies (Santa Cruz Biotechnology, SantaCruz, CA); phospho-S6R, phospho-4EBP1(T37�46), phospho-ACC, phospho-AKT(S473), phospho-p70 S6 kinase(T389), p70 S6

A B

Fig. 7. Evaluation of mTOR activity in FLCN-null and restored cells. (A) Lackof FLCN does not affect mTOR regulation by AMPK. UOK257 (FLCN-null) andUOK257-2 (FLCN-restored) cells were cultured with serum-depleted DMEM for48 h. To activate AMPK, cells were treated with 50 mM 2-deoxyglucose (2DG)for 2 h. AMPK was activated, and mTOR activity was suppressed by 2DG in bothcell lines. (B) Lack of FLCN affects mTOR activity under certain conditions.UOK257 and UOK257-2 cells were cultured without serum for 48 h, stimulatedwith DMEM containing 10% serum, with or without rapamycin (Rapa) (40 nM)or amino acid-free [a.a.(�)] medium with 10% dialyzed serum for 90 min.Under serum-starved conditions, FLCN-null cells did not show complete sup-pression of mTOR activity, whereas FLCN-null cells were more sensitive tomTOR inhibition by amino acid deprivation.

15556 � www.pnas.org�cgi�doi�10.1073�pnas.0603781103 Baba et al.

kinase, phospho-AMPK�(T172), AMPK�, and AMPK�1 antibod-ies (Cell Signaling Technology, Beverly, MA); actin antibody (Bio-medical Technologies, Stoughton, MA); AMPK�1 antibody(Zymed, San Francisco, CA); AMPK� mouse monoclonal anti-body (BD Biosciences, Franklin Lakes, NJ); FNIP-181, rabbitpolyclonal antibody against a FNIP1 synthetic peptide(MAPTLFQKLFSKRTGLGAC); FLCN-102, rabbit polyclonalantibody against a FLCN peptide (QGDGNEDSPGQGEQAEE):FLCN-104, rabbit polyclonal antibody against GST-FLCN; FLCN-mAb from single-clone hybridoma cell line raised against full lengthGST-FLCN in the mouse.

Immunoprecipitation, Western Blotting, and Northern Blotting. Cellswere lysed in lysis or RIPA buffer and immunoprecipitated at 4°Covernight with protein G-Sepharose- (Amersham Pharmacia Bio-tech, Piscataway, NJ) bound antibodies. The resin was washed fivetimes with lysis buffer, proteins were eluted with SDS samplebuffer, followed by boiling, and Western blotting was performed asdescribed (41) (details in Supporting Methods). For Northern blot-ting, a FNIP1 probe was PCR-generated from the unique FNIP1sequence: forward, 5�-TGGAATCATTCAGACCCAGAA-3�; re-verse, 5�-ACATCAGACACTGACGGAACC-3�. A BHD probewas generated by using exon 11 primers (7). Northern blotting wasperformed as described (7) by using a Human 12-lane MultipleTissue Northern Blot (BD Biosciences).

Immunofluorescence Microscopy. Immunofluorescence microscopyof HeLa cells transfected with FNIP1-HA and FLAG-FLCN wasperformed as described in Supporting Methods.

Plasmids, Transfection, Recombinant Protein Expression, and Purifi-cation. The Gateway Protein Expression system (Invitrogen) wasused to produce a variety of expression vectors encoding full-lengthFLCN and FNIP1, FLCN and FNIP1 deletion mutants, and GSTfusion proteins produced in bacmid-infected Sf9 insect cells (detailsin Supporting Methods). Transfections were performed by usingLipofectamine 2000 (Invitrogen). GST-FLCN expressed in Sf9 waspurified on a GST-Trap column (Amersham Pharmacia Biotech).

In Vitro Binding Assay. GST fusion proteins were immobilized onglutathione-Sepharose beads (Amersham Biosciences, Piscataway,

NJ) and incubated with 35S in vitro-translated FLCN or FNIP1prepared by using the TNT T7 Quick kit (Promega, Madison, WI)(details in Supporting Methods). After washing, beads were boiledwith SDS sample buffer and subjected to SDS�PAGE, followed byCoomassie Brilliant blue staining and autoradiography.

AMPK in Vitro Kinase Reactions. To measure AMPK activity inanti-HA-FNIP1 immunoprecipitates, the immunoprecipitates werewashed five times, and the AMPK kinase reaction was performedin reaction buffer with 100 �M SAMS substrate peptide(UpstateBiotechnology, Lake Placid, NY) and [�-32P]ATP (AmershamBiosciences) and assayed according to the manufacturer’s protocol(Upstate Biotechnology). To evaluate HA-FNIP1 phosphoryla-tion, in vitro kinase reactions were performed on HA-FNIP1immunoprecipitates with various kinase inhibitors.

Metabolic Whole-Cell 32P Labeling. HA-FNIP1-inducible HEK293cells were cultured for 24 h in standard medium containingdoxycycline. Cells were washed three times with phosphate-freeDMEM. Cells were labeled for 3 h with phosphate-free DMEMcontaining 10% dialyzed FBS and 1 mCi�ml (1 Ci � 37 GBq)[32P]orthophosphate with and without AMPK inhibitors (AraA, 2mM and Compound C, 30 �M). AntiHA-immunoprecipitatescontaining [32P]HA-FNIP1 were subjected to SDS�PAGE andautoradiography.

We thank Megan Bucheimer, Kelly Esposito, Veronica Roberts, Chris-topher Gunning, and Peter Frank of the Protein Expression Laboratoryfor their excellent technical assistance; Masa-aki Nakaya for help withconfocal microscopy; Debbie Morrison, Akio Yamashita, Tim Veenstra,and Ming Zhou for helpful discussions; and Len Neckers for criticalreading of the manuscript. M.B. received support from the YokohamaFoundation for Advancement of Medical Science. This research wassupported, in part, by the Intramural Research Program of the NationalInstitutes of Health (NIH), National Cancer Institute (NCI), Center forCancer Research. This project has been funded in whole or in part withfederal funds from the NCI, NIH, under contract N01-C0-12400. Thecontent of this publication does not necessarily reflect the views orpolicies of the Department of Health and Human Services, nor doesmention of trade names, commercial products, or organizations implyendorsement by the U.S. Government.

1. Birt AR, Hogg GR, Dube WJ (1977) Arch Dermatol 113:1674–1677.2. Toro JR, Glenn G, Duray P, Darling T, Weirich G, Zbar B, Linehan WM, Turner

ML (1999) Arch Dermatol 135:1195–1202.3. Zbar B, Alvord WG, Glenn G, Turner M, Pavlovich CP, Schmidt L, Walther M,

Choyke P, Weirich G, Hewitt SM, et al. (2002) Cancer Epidemiol Biomarkers Prev11:393–400.

4. Pavlovich CP, Walther MM, Eyler RA, Hewitt SM, Zbar B, Linehan WM, MerinoMJ (2002) Am J Surg Pathol 26:1542–1552.

5. Schmidt LS, Warren MB, Nickerson ML, Weirich G, Matrosova, V., Toro JR, TurnerML, Duray P, Merino M, Hewitt S, et al. (2001) Am J Hum Genet 69:876–882.

6. Khoo SK, Bradley M, Wong FK, Hedblad M-A, Nordenskjold M, Teh B (2001)Oncogene 20:5239–5242.

7. Nickerson ML, Warren MB, Toro JR, Matrosova V, Glenn G, Turner ML, Duray P,Merino M, Choyke P, Pavlovich CP, et al. (2002) Cancer Cell 2:157–164.

8. Khoo SK, Giraud S, Kahnoski K, Chen J, Motorna O, Nickolov R, Binet O, LambertD, Friedel J, Levy R, et al. (2002) J Med Genet 39:906–912.

9. Schmidt LS, Nickerson ML, Warren MB, Glenn GM, Toro JR, Merino MJ, TurnerML, Choyke PL, Sharma N, Peterson J, et al. (2005) Am J Hum Genet 76:1023–1033.

10. Vocke C, Yang Y, Pavlovich CP, Schmidt LS, Nickerson ML, Torres-Cabala CA,Merino MJ, Walther MM, Zbar B, Linehan WM (2005) J Natl Cancer Inst 97:931–935.

11. Warren MS, Torres-Cabala CA, Turner ML, Merino MJ, Matrosova VY, NickersonML, Ma W, Linehan WM, Zbar B, Schmidt LS (2004) Mod Pathol 17:998–1011.

12. Inoki K, Corradetti MN, Guan K-L (2005) Nat Genet 37:19–24.13. Brugarolas J, Kaelin WG, Jr (2004) Cancer Cell 6:7–10.14. Martin DE, Hall MN (2005) Curr Opin Cell Biol 17:158–166.15. Sarbassov DD, Ali SM, Sabatini DM (2005) Curr Opin Cell Biol 17:596–603.16. Hemminki A, Markie D, Tomlinson I, Avizienyte E, Roth S, Loukola A, Bignell G,

Warren W, Aminoff M, Hoglund P, et al. (1998) Nature 391:184–187.17. Jenne DE, Reimann H, Nezu J, Friedel W, Loff S, Jeschke R, Muller O, Back W,

Zimmer M (1998) Nat Genet 18:38–43.

18. Corradetti MN, Inoki K, Bardeesy N, DePinho RA, Guan KL (2004) Genes Dev18:1533–1538.

19. Cheadle JP, Reeve MP, Sampson JR, Kwiatkowski DJ (2000) Hum Genet 107:97–114.20. Eng C (2003) Hum Mutat 22:183–198.21. Harris TE, Lawrence JC, Jr (2003) Sci STKE 2003:re15.22. Fingar DC, Blenis J (2004) Oncogene 23:3151–3171.23. Cook JD, Walker CL (2004) Curr Mol Med 4:813–824.24. Carling D (2004) Trends Biochem Sci 29:18–24.25. Hardie DG (2004) J Cell Sci 117:5479–5487.26. Hardie DG, Hawley SA (2001) BioEssays 23:1112–1119.27. Stein SC, Woods A, Jones NA, Davison MD, Carling D (2000) Biochem J 345:437–443.28. Davies SP, Carling D, Hardie DG (1989) Eur J Biochem 186:123–128.29. Inoki K, Zhu T, Guan KL (2003) Cell 115:577–590.30. Chen J, Hudson E, Chi MM, Chang AS, Moley KH, Hardie DG, Downs SM (2006)

Dev Biol 291:227–238.31. Woods A, Johnstone SR, Dickerson K, Leiper FC, Fryer LG, Neumann D, Schlattner

U, Wallimann T, Carlson M, Carling D (2003) Curr Biol 13:2004–2008.32. Shaw RJ, Kosmatka M, Bardeesy N, Hurley RL, Witters LA, DePinho RA, Cantley

LC (2004) Proc Natl Acad Sci USA 101:3329–3335.33. Inoki K, Li Y, Xu T, Guan KL (2003) Genes Dev 17:1829–1834.34. Shaw RJ, Bardeesy N, Manning BD, Lopez L, Kosmatka M, DePinho RA, Cantley

LC (2004) Cancer Cell 6:91–99.35. Cantley LC, Neel BG (1999) Proc Natl Acad Sci USA 96:4240–4245.36. Leevers SJ, Vanhaesebroeck B, Waterfield MD (1999) Curr Opin Cell Biol 11:219–

225.37. Inoki K, Li Y, Zhu T, Wu J, Guan KL (2002) Nat Cell Biol 4:648–657.38. Potter CJ, Pedraza LG, Xu T (2002) Nat Cell Biol 4:658–665.39. Iwamatsu A (1992) Electrophoresis 13:142–147.40. Jensen ON, Podtelejnikov A, Mann M (1996) Rapid Commun Mass Spectrom

10:1371–1378.41. Baba M, Hirai S, Yamada-Okabe H, Hamada K, Tabuchi H, Kobayashi K, Kondo K,

Yoshida M, Yamashita A, Kishida T, et al. (2003) Oncogene 22:2728–2738.

Baba et al. PNAS � October 17, 2006 � vol. 103 � no. 42 � 15557

MED

ICA

LSC

IEN

CES